Voice Audit

A voice audit is the process of collecting your real existing copy across every surface then dissecting it line by line to reverse engineer the actual personality already living in your best work. You pull homepage headlines from 2024, error messages from your app, welcome emails, support docs, social posts, checkout flows, and even old founder letters. You line them up and ask hard questions. Which sentences feel like they could only come from your team? Which ones could be swapped with Notion or Stripe or any random SaaS tool and nobody would notice? The output is not a mood board. It is a concrete set of rules extracted from your own history that tells writers and AI generators exactly what your voice permits and forbids. Mailchimp did versions of this before they published their famous guide. They looked at their early email campaigns and saw that usefulness always outranked cuteness so they codified it. The audit becomes the foundation for voice attributes, the never say list, and the before and after rewrites that actually train humans and machines alike.

A voice audit is not a creative workshop where you invent a personality over oat milk lattes. It is not hiring an agency to audit your competitors and then bolt their best traits onto your deck. It is not a vibe check or an archetype quiz or a session where everyone votes on whether your brand is more like Discord or Headspace. Those exercises produce decoration. A real audit starts with the copy you have already shipped and forces you to confront the gap between what you think you sound like and what actually lands in the world. It does not let you hide behind aspirational language. If your billing error copy reads like a 1998 IBM mainframe while your Instagram sounds like a hypebeast, the audit calls that out without mercy.

Look at what Hey did in 2020. Basecamp collected every scrap of copy from their old products plus the first drafts of the Hey landing page. They found their highest confidence voice lived in the blunt founder emails Jason Fried wrote at 6am. Sentences like we built this because the other stuff sucks. The audit revealed that hedging language like we are pleased to offer appeared in legal flows and support docs but never in the copy that converted users. They extracted three rules: name the enemy out loud, use active voice that assigns blame to broken systems, and never soften a strong opinion with maybe or perhaps. The concrete example that emerged was rewriting a generic modal from are you sure you want to delete this conversation into delete it. Good. Those pixels were spying on you anyway. That single rewrite became the template for every UI string that followed. The audit took two days and gave them a voice that positioned the product as a weapon against surveillance capitalism instead of just another email client.

Innocent Drinks ran a similar audit in the UK years before they scaled to supermarket fridges. They gathered every carton side panel, every early website line, every customer service reply from their first three years. The pattern that jumped out was radical honesty wrapped in a dad joke. When a batch went wrong they did not say we apologize for any inconvenience caused. They said our strawberries had a bad week. Sorry about that. The audit produced a four beat pattern: greeting, small fact, small joke, treat you like a human. They used it to rewrite a delayed delivery email from our logistics partner experienced an unforeseen delay into the van broke down and the bananas are now having a party in the road. That concrete rewrite trained every new writer and kept the voice intact even as the company grew to hundreds of employees.

Run a voice audit the moment your team grows beyond one writer or when you plan to feed copy to AI generators. Do it before a product launch that will generate dozens of new error states and onboarding flows. Use it when your homepage sounds confident but your support documentation reads like it was written by a committee of lawyers on sedatives. It is the fastest way to fix the slow drift that happens when marketing sounds like a startup, legal sounds like a law firm, and product sounds like whatever engineer was least busy that Friday. Skip the audit if you are a solo founder who still writes every word and it already feels consistent. Do not run one in the middle of a full rebrand when the product promise itself is changing because you will be auditing a ghost. And avoid it if your leadership team cannot stomach owning a distinct point of view. The audit will expose that you have nothing sharp to say and that realization tends to sting.

Discord performed an informal audit when they expanded beyond gaming. They collected voice chat error messages, server invite copy, and community guidelines. The audit showed their best work sounded like a helpful friend at 2am not a corporate community manager. So they banned phrases like we are committed to fostering positive interactions and replaced them with chill, this server is what you make it. The concrete examples trained their AI tools to stop generating stiff moderation messages and start generating ones that actually match how their users talk to each other.

Headspace used an audit to protect their calm deliberate voice when they expanded into enterprise wellness. They pulled meditation reminder copy, app store descriptions, and billing emails. The audit flagged places where urgency crept in during error states. A push notification that once read your session is overdue became you have time. Come back when you are ready. That single distinction kept the product experience consistent even as the brand scaled.

The audit forces you to write the never say list from reality instead of theory. If your collected copy shows three different versions of we leverage cutting edge technology then utilize leverage and cutting edge all go on the banned list with plain replacements that match your actual best work. It produces the before and after rewrites that become training data. A weak sentence like our end to end platform maximizes synergistic outcomes gets placed next to it handles the heavy lifting so you can focus on what matters. Writers and AI models learn faster from these real pairs than from any adjective list.

Most brands treat voice as something they add later. The audit flips that. It proves the voice already exists in fragments. Your job is to find the strongest fragments, kill the imposters, and turn the survivors into law. Do this once properly and every future piece of copy improves because the target stops moving.

A voice audit turns the copy you already shipped into the only blueprint you will ever need.